Time Series Prediction of AQI in Shenzhen Based on ARIMA Model

Objective To establish appropriate prediction model of Air Quality Index(AQI) in Shenzhen on autoregressive integrated moving average (ARIMA) model, and provide a scientific basis for control of air pollution.

Methods Time series analysis was conducted by using the daily data of AQI in Shenzhen from January 1, 2014 to June 30, 2016, and a predictive model was established after parameter estimation, model diagnosis and model evaluation.The optimal prediction model was selected. The model was used to predict the value of AQI from July 1, 2016 to July 6, 2016, and the prediction effect was evaluated.

Results 912 daily AQI values were collected from January 2014 to June 2016 in Shenzhen. The proportion of air quality levels for optimal, good and mild pollution were 48.6%, 48.4% and 48.6% respectively. Through the test of stationarity, ARIMA(3, 0, 1) was selected as the optimal model. The AIC and BIC of this model were 7 364.51 and 7 393.41 respectively, which were the least. The Q statistic was 17.48 (P>0.05) by Box-Ljung testing, indicating the applicability of the model. The average relative error between the predictive value and the actual value of AQI from July 1, 2016 to July 6, 2016 was 16.6%.The actual values were within 95% CI of the predictive values. The established ARIMA(3, 0, 1) model was good in fitting precision and prediction effect.

Conclusion The ARIMA(3, 0, 1) model could predict the change trend of AQI in Shenzhen with good prediction effect.